John Galt Talks Pixels

[David Birdy]

I love the tech articles and theories.. (even when they are written with an air of marketing) it's really half the fun of this brave new digital world!

The fact that we all have eyes and can see an image, is the flip side of the technical theory.... if the image looks good, then it's a good image!

 

[Jacek Zakowicz]

Optimator tests

Optimator tests

When you look at a zone plate from the F35 / Genesis you'll see a few things:

1) different resolution horizontally than vertically
2) more luma moire vertically
3) chroma moire / rainbows horizontally
4) overall resolution sub 1920 lines (about 75% of that visible in image projection)

So how does that answer your question? Let's compare with the RED One:

1) same resolution horizontal and vertical
2) negligible luma moire
3) negligible chroma moire
4) overall resolution vastly greater than 1920 lines. Approx 3.2k visible on projection. MTF is very strong so that unlike the F35, 2k resolution is strongly visible and detailed with strong MTF.

Theory becomes meaningless when you build the camera and shoot the image.

Graeme

About 30 seconds with the Optimator will clearly show that your point #4 on the F35 and point #1 on the Red One are incorrect.
Not trying to stir just stating the facts. I've tested both cameras numerous times....:thumbup1:

 

[Kwan Khan]

#1 on the Red One are incorrect.
Not trying to stir just stating the facts. I've tested both cameras numerous times....:thumbup1:

Please provide with more details of this operation (will be appreciated).

 

[David (dudi) Namir]

But what? The proof is in the pudding!

Graeme

Have two Reds and going on stage two probably with both, but i do believe the correct phrase is:
"The proof of the pudding is in the eating."
All is good Graeme.

 

[Graeme Nattress]

If you want to get into the details, on the F35 the vertical resolution, in terms of camera response to input detail is greater than 1080 because of the aliasing. The camera responds and produces image for such input, but the image is corrupt through aliasing. I don't see how that can count as resolution. Similarly in the horizontal direction where there is a unique pixel for each of the 1920, but as the detail gets fine enough you get a rainbow of redish, greenish, blueish pixels in order. The eye integrates this into grey from a distance. When you zoom in, you see the false colours. Again, hard to count as resolution.

Whenever I've measured the RED One, it's shown the same resolution horizontally and vertically, which is plainly not the case on the F35. We've posted the images before that show this.

Graeme

 

[Hrvoje Simic]

Nice Panavision propaganda piece.

A number of years ago some IMAX engineers – and I don’t think IMAX ever let these guys out of their lab again -- did this wonderfully elegant experiment at the Large Film Format Seminar at Universal Studios Imax theatre. They showed this film they made that began with 2 rows of 2 squares: black white, white black, as if you had 4 pixels on the screen.

Then they started to double and double and double the squares. Before they got to 4K the screen was gray. Do you know what the means? There was no longer any difference between black and white, which is what allows you to see sharpness. It's the contrast that we see, not the actual information. Technically, the MTF (Modulation Transfer Function) was zero at 4K!

Let's just pretend for a moment that IMAX truly is 4K. You watch IMAX at between one and one and a half picture heights from the screen. But in order to get to appreciate 4K on a regular movie screen, you would have to sit much closer than normal. In other words, when you go to a movie theater, and most of the modern theaters with stadium seating are designed so that the middle of the theater is 2 ½ to 3 picture heights from the screen, for most of us who watch movies, that’s pretty where we want to be sitting. Maybe just a little bit closer from some of us who do this for a living, because we're maybe looking for artifacts or issues. If you sit much closer than 2 ½ picture heights, that's what you're seeing, artifacts, not movies!

So if you had true 4K resolution in your local theater, everybody would have to sitting in the first 6 rows. Otherwise they wouldn't see any extra detail. Their eyes wouldn't LET them see it. You know this intuitively from passing by these beautiful new monitors at trade shows. You find yourself getting absolutely as close as possible to see the detail, and to see if there are any visible artifacts. At normal viewing distances, you can't.

Very nicely packaged information to achieve a goal.
Too bad it is based on a false premise and actually stating complete nonsense.

"Perceivable resolution based on recognizing individual rows or pixels" is a complete bullshit theory.


1. Eye doesn't have to see an individual pixel in order for the picture to become more clearer, The difference in perception exists anyway and is obvious, depending on the properties of the medium which is showing the imagery. After visually quantifiable difference perception of clarity reaches further.

The same principle applies in colour depth perception as the eye doesn't have to perceive a difference between 16M or 1B individual colour tones in order to preceive more natural overall color of the image. Even without seeing "banding", which is a display technology artifact.

People who still buy "no extra detail perceived" nonsense stated in that quote might want to pay a visit to CES 2011.

2. Exploring and defining human vision, an organic sense, strictly by using pixels and squares is ridiculous.

3. If anyone wants to offer more "4K is too much for you to see" propaganda they are welcome to substantiate statements with showing stuff like grass, face and cloth instead of unified patterns. In 1080p, 4K and 8K. Side by side. On a same screen size and distance. With latest display technologies. No need to bring a calculator or ruler. Expression on faces of the audience is enough.

4. Stating that human eye doesn't recognize the difference beyond 1080p per cinema FOV is an insult to human vision.

5. Using monitor example to base a statement about capabilities of human vision is based on a premise that monitors have picture quality representation equivalent to human vision. Also - ridiculous.
Monitor example also shows that the author of the statement is not familiar with reflection, refraction, diffusion, dynamic range, pixel distance, refresh rate, etc...etc...etc...or just assumes that the ones to whom statement is directed to aren't.


Ones who state this nonsense are either completely ignorant about the matter they are covering, or willingly misleading people in order to make a buck. It's one thing to devaluate competition's product just to sell your own, that's just lame, but to go so far and undermine human capabilities...that's really low.





For those who like to play with numbers:

Quality print would be at least 300 DPI.
It looks pretty good and people don't look for pixels or black and white squares.

Cinema theatre FOV at arm length distance would be roughly 80 cm in width - for those who like to sit more at the back,
and roughly 100 cm in width - for those who prefer middle seats.

 

[Graeme Nattress]

It's plainly obvious to anyone who's had the visit to the RED Studio cinema that the best "2k" camera out there - the F35 doesn't have the resolution or MTF of even the RED One, and that's in our 4k projection environment, not pixel-peeping. Similarly we've seen 4k as 4k and 4k downsampled to 2k and you can see the resolution improvement in the 4k over the 2k.

Graeme

 

[Jacek Zakowicz]

clarification

clarification

It's plainly obvious to anyone who's had the visit to the RED Studio cinema that the best "2k" camera out there - the F35 doesn't have the resolution or MTF of even the RED One, and that's in our 4k projection environment, not pixel-peeping. Similarly we've seen 4k as 4k and 4k downsampled to 2k and you can see the resolution improvement in the 4k over the 2k.

Graeme

Greame,
I'm not disputing that RedOne has more resolution than F35- It does for sure.
By a lot...
But the rolling shutter (I think ) introduces "jaggies" artifacts close to the treshold frequency on the Redone and my vertical and horizontal trumpets resolve differently. of course that's the live view "dirty" debayer- I understand it-but it's hard to believe that it would not affect R3Ds.,
now I can tell you that F35 resolves fine into the 1920x1080 frequency on the Optimator target. Both horizontal (that shows aliasing) and vertical.
Now that's all live view out of the camera on monitor.
No recording, compression, playback, etc...
I think that's a pretty fair Sensor test- no lenses, back focus, lighting variations involved.

 

[Nick Shaw]

…of course that's the live view "dirty" debayer…

I don't think looking at the 720p live output of the RED One is a very good way to judge much at all. And the 4k does not even fill the 720p due to the look around. I wouldn't say it's quite like judging a film camera by the video tap, but…

The live view from the Genesis/F35 is the best image quality attainable by the camera. The live view of the RED One is not. Not apples to apples.

 

[Jacek Zakowicz]

I don't think looking at the 720p live output of the RED One is a very good way to judge much at all. And the 4k does not even fill the 720p due to the look around. I wouldn't say it's quite like judging a film camera by the video tap, but…

The live view from the Genesis/F35 is the best image quality attainable by the camera. The live view of the RED One is not. Not apples to apples.

This is correct but your assumption that I looked at 720p is not (correct)
You can zoom in 1:1 remember? in live view. I only need to see the center of the sensor since the pixels are uniform throughout...:idea:
Just to be clear the treshold frequency for RedOne is much higher than that for F35 and by the time you get down to 1080p (on the Redone) you should be able to easily remove any (resolution related) artifacting from the M or MX sensor

 

[Patrick Tresch]

You can zoom in 1:1 remember? in live view. I only need to see the center of the sensor since the pixels are uniform throughout...:idea:

Not sure if this was designed to calculate resolustion...

Patrick

 

[Nick Shaw]

This is correct but your assumption that I looked at 720p is not (correct)
You can zoom in 1:1 remember? in live view. I only need to see the center of the sensor since the pixels are uniform throughout...:idea:

Fair enough on my first point. My second still holds true though.

 

[Jacek Zakowicz]

will check R3Ds when I can

will check R3Ds when I can

Fair enough on my first point. My second still holds true though.

OK next opportunity I'll record and go through full R3d debayer. I don't expect it to be much different. After all it's only a black and white high contrast stationary image. Pure resolution test......

 

[Graeme Nattress]

The real time demosaic in the R1 could indeed produce a differential resolution horizontally than vertically. It certainly doesn't carry through to the recording. If you were judging off the live feed, I can see where you're coming from. I'm always looking at the RAW R3D recording - in other words, testing the whole chain, lens included.

 

[Martin Stevens]

Wow! I just read the John Galt article and found it to be full of conclusions that are not at all accurate.

He seems to deduce things based on abstracts and not reality.

Graeme and Red are correct, because you can empirically validate all of it with your eyes.
Seeing is believing. And I'm not talking about emotional reactions to the images either.

There is no reason to FEAR bayer pattern sensors.

 

[Deanan]

OK next opportunity I'll record and go through full R3d debayer. I don't expect it to be much different. After all it's only a black and white high contrast stationary image. Pure resolution test......

You don't expect a full resolution debayer to be different in resolution to be different than a quick and dirty downsample? Quite surprising coming from someone as technically adept as you.

 

[Deanan]

Very nicely packaged information to achieve a goal.

Too bad it is based on a false premise and actually stating complete nonsense.

"Perceivable resolution based on recognizing individual rows or pixels" is a complete bullshit theory.


1. Eye doesn't have to see an individual pixel in order for the picture to become more clearer. The difference in perception exists anyway and is obvious, depending on the quality of the medium which is showing the imagery. After quantifiable difference clarity reaches further. People who still buy that nonsense might want to pay a visit to CES 2011.

2. Exploring and defining human vision, an organic sense, strictly with pixels and squares is ridiculous.

3. If anyone wants to offer more "4K is too much" propaganda let's see grass, face and cloth instead of unified pattern. No need to bring a calculator or ruler.

4. Stating that human eye doesn't recognize the difference beyond 1080p per cinema FOV is an insult to human vision.

Agreed. The eye can pick out detail differences even at long distances and less than perfect vision.

2k was the baseline to be acceptable to compare to a film print which is 1.2-1.5k but has grain to help add perceptual sharpness. How it went
from minumum required to all that you'll ever need is all "marketing pixels" too .

 

[Jacek Zakowicz]

Downsample

Downsample

You don't expect a full resolution debayer to be different in resolution to be different than a quick and dirty downsample? Quite surprising coming from someone as technically adept as you.

Deanan,
I would expect the 1:1(pixel for pixel, pixel-mapped-hint, hint) live view to be called a lot of things but a down sample by someone as technically adept as yourself:reddevil:
On a serious note I will rephrase it:
I will make no assumptions as for the outcome- I'll just record/full debayer the nearest opportunity and compare.:thumbsup:
Anyway, as I said-this has no practical effect on the product(footage)- I was just pointing out my observation....

 

[Deanan]

Deanan,
I would expect the 1:1(pixel for pixel, pixel-mapped-hint, hint) live view to be called a lot of things but a down sample by someone as technically adept as yourself:reddevil:
On a serious note I will rephrase it:
I will make no assumptions as for the outcome- I'll just record/full debayer the nearest opportunity and compare.:thumbsup:
Anyway, as I said-this has no practical effect on the product(footage)- I was just pointing out my observation....

OK, I could have said debayer/downsample

1:1 is a quick and dirty 'debayer' and not appropriate for for making the kind of conclusions you're making. It's very much like looking at video tap off a film camera and concluding that the resolution isn't there.

 

[Sanjin Jukic]

Whole point of this tread is that Jim Jannard did it and John Galt didn't.

Then John Galt took opportunity to speak loud what he thinks exactly about the thing that he couldn't make it actually...

But those are almost always totally two very different things: one is about theory and another is pure reality.

If you open the last issue of American Cinematographer then you can read that even Clint Eastwood

as a hardcore Panavision anamorphics guy using RED1-MX for his VFX shots at his last movie.

Very soon Epic and Scarlet are reality...